Hi Paul, Paul Gevers, on 2024-10-21: > On Mon, 21 Oct 2024 21:38:40 +0200 =?utf-8?Q?=C3=89tienne?= Mollier > <emoll...@debian.org> wrote: > > While staring at the ongoing perl 5.40 transition, we noted that > > on rare occasions[1], > > It's not rare. Since the start of the perl transition, no run has passed on > armhf.
Uuh, right, indeed I also reproduced the bug locally in a qemu armhf build environment. > > the debian/test/gui autopkgtest script > > could fail, presumably with the Xfvb server not having finished > > to start up yet when the command to be tested attempts to spawn > > a gui, resulting in: > > And this is what I see when I run Xfvb in the testbed where the test just > failed: > root@elbrus:/tmp/autopkgtest-lxc.ss1_xnf2/downtmp/build.ZNN/src# Xvfb :5 > (EE) > (EE) Backtrace: > (EE) > (EE) Segmentation fault at address 0x8040201 > (EE) > Fatal server error: > (EE) Caught signal 11 (Segmentation fault). Server aborting > (EE) > Aborted The issue looks very specific to this way of running Xvfb :5, other ways I know of within autopkgtests seem to be running fine in the same configuration. Quick lookup of open xvfb issues suggest that #1084230 could be relevant to the problem at play, and might also explain why there is not a large fallout of issues in all tests depending on xvfb, but just ponctual problems related to a particular way to launch the X server. Thank you for correcting me and the complement of information! Have a nice day, :) -- .''`. Étienne Mollier <emoll...@debian.org> : :' : pgp: 8f91 b227 c7d6 f2b1 948c 8236 793c f67e 8f0d 11da `. `' sent from /dev/pts/2, please excuse my verbosity `- on air: Presto Ballet - Find The Time
signature.asc
Description: PGP signature
